Empower people through tech. Everyone can build their automation. Engineer, farmer. Live in a Himalayan village. https://xitter.cf/t.co/RvU1niKIbE
Himalayan village, India
Joined March 2008
- Tweets2.7K
- Following589
- Followers917
- Likes2.3K
Pinned Tweet
Sometimes it takes years for an idea to work out.
I own my product's domain since 2013! Chased and failed multiple times. It was beyond what a solo founder could do.
Then came LLMs, they got better at code. I restarted my product. Fresh perspective and a decade of attempts.
For most software, there is a ceiling - fit your real life operations.
A business/personal software should not be imaginary - that becomes useless.
Have a real world use-case and build that fast and make it fit tight.
Reach that level, then stay (maintain + minor additions).
Just installed github.com/indexable-inc/bla… and will give it a try with my active projects.
Building with LLMs, Rust has been a massive win, but the long build times bite.
Maybe this is now solved? If active project builds in milliseconds, it will surely improve my workflow.
Sumit Datta retweeted
384 bits is 48 bytes. Pick a random number between 0 and 255. Now do that 47 more times. What are the odds you picked 00 for the first 47 picks?
Sites built with PHP are feeling sorry.
And a few PHP bros showing how quickly they patched everything.
Sell shovels.
World needs a better alternative.
It is funny when a VC points this out when it is literally our financial incentives and culture around finance that makes these happen.
Again and again.
Gross incompetency from Indian certificate authority eMudhra.
You had one job: generate a 384 bit private keys and instead you used: 4, 56, 66, 115, 155, 156, 230, and 237.
This allows attackers to issue certificates for any websites.
Embarrassing that it’s a family business run by largely computer illiterate people. Only 1 of the 3 have a remotely computer-related background and even they never actually worked a technical job in their life. If you look at the LinkedIn, their tech team is young, from tier 3 education and paid sort of bottom barrel.
Imagine how many attacks you can design with AI given the existence of this genre of gross negligence.
A lot of well meaning folks are trying to say software is not dead.
Do NOT fall for this.
We are in uncharted territory.
"Wait and watch" should be the right option for millions.
If this shit goes south, it will really hurt people. A lot of people. Be cautious.
In 1994, software was dead. People told me programming was for nerds and that I should become a doctor or a lawyer instead.
In 2004, software was dead. People told me that someone from India would take my job for $5/hour.
In 2014, software was dead. People told me that no-code was going to kill every job and doom my career.
In 2024, software was dead with Devin, Windsurf, Copilot, and OpenAI o1.
We kill software every 10 years, and yet the demand for good Software Engineers has never been higher.
I want LLM generated alternative to Android, the OS.
I want to use my Android (compatible) phones with a different OS, one that is built with less clutter, low system resource usage.
A customizable build that I can load on a phone would be awesome!
LLM code to generate finite state machines as per user's intent.
Separate machines for backend, user auth/permissions, frontend...
Perhaps that is the future of software.
Who is going to tell this person that it is October 2026?
Things change, a lot, every, month.
No hay nada que me diga que esto es cierto.
Hoy se manda muchas cagadas cualquier modelo.
Entonces no se la verdad si se quieren sumar a una movida o yo estoy haciendo algo muy diferente.
No se ni que significa que el software esté muerto. Todo lo contrario si un modelo puede ser un equipo entero de ingenieros debería todos estar haciendo lo que necesitan super personalizable. Lo cual no siempre es bueno.
Good agent workflow is mostly chat manipulation:
1. LLM shared questions? Split and dig into each - multiple forks of chat, merge, resume
2. I have questions about current chat - Fork that, dig in, join parent, resume
3. Multiple items - fanout (impl, review), merge
Every time I have tried to build a better UX for coding agent, where I can see the code that is planned for change or being changed, I get the feeling this is a lost cause.
We need better ways to track change which other agents can review. We cannot review changes visually.
I do not think even engineers really love computers. They love their paychecks and did "computer" work.
LLMs are basically math. It cannot get more "computer". And still engineers want to ban them.
This sudden obsession with "read the code" is meaningless. Most engineering teams happily avoided reading code from their own team.
Frontend, backend, devops, sysops, db developer, API engineer... No one read the others' code.
Now that LLMs write code, people are going nuts.
This.
This is what will happen to software.
Written, re-written, ported or built from scratch.
All by LLMs. No human drama needed.
Announcing tsc-rs (aka ts-rust), my complete rewrite of the Typescript compiler, type checker, and LSP, all in Rust.
This is a project I've had agents working on for 5 months now. I burned ~$400k of Codex tokens and got nowhere. Burned ~$20k of Opus and got there in 2 weeks.
I have not read a single line of the code.
tsc-rs is an open source drop-in replacement for tsc, and it's available now.
Every line of software that can be written by LLMs, will be written by them.
Written, re-written, ported and build in ways that most software engineers of today will have a hard time believing.
But it will happen.
"One project, one harness":
Projects fall into a pattern after some development - tech stack, scripts, deployment, etc.
Instead of Markdown files, what if harness grew with a project?
Agent + tools + permissions for planner (Epics/Tasks), frontend, backend, deployment...
Sumit Datta retweeted
a weird inversion with LLMs is the models improve faster than the tinkerers
when i see people with custom workflows and setups they're all addressing problems that don't exist anymore
the person naively using vanilla codex is more likely to be experiencing state of the art
I have been doing "Code distillation".
Use a capable LLM to discuss how your want to process a file or instruct some automation.
Then ask agents to write tests, build a parser. Really large and lenient parser.
For example: CSV processor with simple language, no AI at runtime.
Sumit Datta retweeted
Orders are now open for the new Lucebox Zero 495: shop.lucebox.com/
It’s AMD’s new Ryzen AI MAX+ PRO 495 with up to 192GB of unified memory, plus a 32GB Radeon AI PRO R9700 in the same box.
Lucebox engine split the model by how often each part runs. Dense layers, the hottest experts and the drafter go on the R9700, because nearly every token reads them. The long tail of experts, which most tokens skip, stays in the 495's unified memory. Both chips compute at the same time. 15 tok/s on the 495 alone, 50 with the R9700.
You can get also up to 4TB SSD and 100 GbE cluster kit, installed.
First 200 units are $1,000 off. Deliveries expected in January.
